SLA & Rate Limits

Uptime. Latency. Limits.

99.9% monthly uptime commitment on paid tiers. 99.97% trailing 90-day actual. p95 memory-search latency under 180ms globally. Credits automatic when we miss.

Uptime

99.9% monthly. 99.97% actual. Credits if we miss.

Commitment

99.9%

Monthly uptime on Pro, Team, and Enterprise tiers (99.95% on Enterprise dedicated).

Trailing 90-day actual

99.97%

Last-90-days uptime across the memory API. Live numbers at /status.

Service credits

Automatic

Pro-rated credits applied to your next invoice if monthly uptime drops below commitment. No claim needed.

Credit schedule: <99.9% but ≥99.0% → 10% of monthly fee credited. <99.0% but ≥95.0% → 25%. <95.0% → 50%. Planned maintenance (with 48h notice) and customer-caused incidents are excluded. Full definitions in the signed DPA/MSA.

Latency

Fast enough to feel instant.

memory-search-semantic · p50

<50ms

Median latency measured from US-East. Rolling 30-day window.

memory-search-semantic · p95

<180ms

95th percentile target. Our 90-day actual: 142ms.

memory-set (write) · p95

<120ms

Includes embedding, vector index update, and FTS index write.

Live latency distribution at /status. Self-hosted deployments depend on your hardware — see the benchmarks page for reference numbers.

Rate limits by tier

Predictable ceilings. Burst-friendly.

Limits are enforced at the API gateway. Exceeding a limit returns 429 Too Many Requests with a Retry-After header. Short bursts (up to 5x sustained rate for 10 seconds) are allowed.

Tier	Price	Daily requests	Burst / minute	Concurrent
Free	$0	100 / day	20 / min	2
Pro	$20 / mo	10,000 / day	200 / min	10
Team	$99 / mo	100,000 / day	1,000 / min	50
Enterprise	Custom	Negotiated (no cap)	Dedicated capacity	Dedicated capacity

Dream Engine cycles, background consolidation jobs, and integration syncs count toward daily requests at discounted rates — full schedule in the API docs.

Fair use policy

Designed for real products, not abuse.

Tier limits apply per account, not per API key. The following are considered abuse and may result in throttling, account suspension, or tier re-classification:

Creating multiple free accounts to circumvent tier limits.
Running unattended scrapers or crawlers against our infrastructure.
Using the service as a bulk storage layer for content unrelated to memory / AI agent context.
Sustained usage patterns more than 10x your tier's daily cap for 3+ consecutive days (we will reach out before taking action).

If you're building something that stresses these limits for a legitimate reason, email legal@remlabs.ai before you hit production and we'll size you on enterprise.

Questions?

Need a custom SLA or BAA?

Enterprise customers get financially-backed uptime and latency SLAs with dedicated capacity, custom metrics, and a signed agreement.

Talk to sales View live status → Security overview